I turn ambiguous, high-stakes operating workflows into reliable AI products—from discovery and architecture through APIs, evaluation, deployment, and adoption.
My background combines hands-on agent building with live commercial operations, private-wealth product delivery, fund diligence, and credit-risk analytics.
A closed-source production case study showing how I make a non-deterministic multimodal agent accountable through evidence contracts, evaluation gates, job isolation, scoped recovery, human authority, and operational readback.
Read the GitHub case study · View the visual walkthrough
- Agent and workflow/state architectures
- Evaluation, guardrails, observability, rollback, and recovery
- Human-in-the-loop operational products
- Data reconciliation and decision-support systems
- Client-facing AI solutions from ambiguous requirements to adoption
Hong Kong permanent resident · Open to APAC and international opportunities